Learning Health Systems provide a glide path to safe landing for AI in health

Area of Science:

• Healthcare Informatics
• Artificial Intelligence in Medicine
• Health Systems Science

Background:

• Artificial Intelligence (AI) shows great potential in healthcare but faces challenges in clinical integration.
• Current AI deployment models often lack the infrastructure for sustained, adaptive use in complex healthcare settings.

Purpose of the Study:

• To propose Learning Health Systems (LHS) as a socio-technical framework for the safe and sustainable integration of AI into healthcare.
• To outline how LHS can address key challenges in AI implementation, including generalizability, workflow integration, and ethical considerations.

Main Methods:

• Conceptual framework development based on insights from an international workshop (2025).
• Analysis of LHS principles (co-creation, real-world evaluation, continuous learning, governance) in the context of AI deployment.
• Discussion of practical steps for aligning AI development with LHS frameworks.

Main Results:

• LHS provide a 'glide path' for AI integration by enabling iterative development, monitoring, and recalibration.
• LHS embed essential functions for AI governance, including accountability, data provenance, and adherence to FAIR principles.
• LHS facilitate 'human-in-the-loop' safety mechanisms and promote a culture of continuous learning around AI tools.

Conclusions:

• Integrating AI within LHS transforms implementation into a sustained, evidence-based learning process.
• This approach aligns AI innovation with clinical realities, enhancing patient care, health equity, and system resilience.
• LHS offer a strategic vision for the future of AI in healthcare, ensuring responsible and effective adoption.